Research And Implementation Of Echo Cancellation Based On NLMS Algorithm

Most hearing impaired patients rely on digital hearing aids to compensate the hearing. The quality of hearing aids on the market varies greatly, and a major aspect of the performance difference is the echo cancellation in a hearing aid. Echo has a great influence on the speech intelligibility and comfort. In order to solve the problem caused by echo in the hearing aid, this paper researches and realizes the application of adaptive algorithm in echo cancellation of digital hearing aids.The Least Mean Square(LMS) algorithm is often used in echo cancellation.However, due to the slow convergence speed and large steady-state error, LMS can't meet the requirements of instantaneity and clarity of hearing aids. Therefore, the Normalized Least Mean Square(NLMS) algorithm is adopted in this paper when the update frequency and step size are adjusted.In this paper, the algorithm of echo cancellation of digital hearing aid contains:echo detection algorithm, improved NLMS algorithm, echo delay estimation algorithm and Non Linear Processing(NLP) algorithm, in order to achieve the accuracy and instantaneity of processing of speech signal. At the entrance of the algorithm module,the echo detection device is added. When the echo energy is lower than the threshold,the device can avoid the frequent jump of the filter and save the system cost.Improvements on the NLMS include: updating filter coefficients by batch and joining the factors to control iteration step size. In this way, improved NLMS reduces the computational complexity and gets faster convergence speed. Echo time delay algorithm is used to estimate the delay between the processed speech signal and the desired signal. The two buffer arrays can be aligned to ensure the correct processing of data by the delay. The algorithm computes the cross-correlation of the desired signal and the processed signal, and computes the cross-correlation of the results and autocorrelation of the desired signal. Time delay estimation through two cross-correlation has stronger anti-noise ability and stability. NLP algorithm uses a filter based on amplitude to prevent low-rising value signal and allow high amplitude signal. Through this processing, residual echo can get further eliminated, which makes the voice output have better comfort.The improved NLMS algorithm is simulated on MATLAB to analyze the performance comparison with the original NLMS. Then analyze and compare the processed voice with other algorithms by the Cool EditPro. The simulation results of the improved algorithm are better than the traditional algorithms. The improved NLMS has high convergence speed, smaller echo residual and better comfort.